skills.rest
Nitrogen216
Nitrogen216COMMUNITY
Explore all skills by this author →
GitHub

meta-apply

Applies staged self-modification patches to a skill corpus after cross-model jury review and human approval.

Updated Aug 18, 2026
Search Tags:#code review#provenance#audit trail#agent governance#self-modification#patch apply#cross-model jury
One-click install
npx skills add https://github.com/Nitrogen216/awesome-dsh-mods --skill meta-apply-nitrogen216
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: meta-apply
Source: https://github.com/Nitrogen216/awesome-dsh-mods/tree/main/modes/dsh-autoresearch/vendor/aris-upstream/skills/meta-apply
Command: npx skills add https://github.com/Nitrogen216/awesome-dsh-mods --skill meta-apply-nitrogen216

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Self-modifying agent systems risk applying their own proposed changes without oversight. This Skill enforces a privilege boundary: producer skills can only stage candidate patches, and only this human-invoked applier can land them, with a fresh cross-model jury verdict obtained at landing time. ## Core Features & Use Cases - Jury-at-landing verification: Runs a fresh cross-model review (via codex MCP) on each staged diff at apply time, ignoring any producer-written verdict that could be forged. - Provenance stamping: Records author, reviewer, verdict ID, and content hash on every landed change via a provenance helper, with cross-family author/reviewer enforcement. - Reversible mutations: Backs up every target file before applying patches and logs each landing to an audit trail. - Use Case: After running a meta-optimization pass that staged three candidate patches, invoke this Skill to jury-review and land only the patches you explicitly approve. ## Quick Start After reviewing the staged patches from a meta-optimize run, ask the agent to run meta-apply on the patch numbers you approve, such as patches 1 and 3.

Frequently Asked Questions about meta-apply

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I apply staged patches from a meta-optimize run?▼

Invoke meta-apply with the patch numbers you want to land, such as /meta-apply 1,3 or all. The skill runs a fresh cross-model jury on each staged diff and only lands patches that pass and were explicitly named.

Why does meta-apply re-run the jury instead of trusting the producer verdict?▼

A producer-written jury_verdict field is forgeable, since the producer writes it. The applier runs its own fresh cross-model review on the actual staged diff at landing time, so the verdict is produced where it is consumed.

Can a human override a jury KILL verdict?▼

No. The human can only select among patches the jury passed; a KILL verdict permanently refuses that patch. This keeps the reject-default rule structural rather than advisory.

What happens if the author and reviewer are from the same model family?▼

The provenance helper's assert_cross_family check raises and the patch is refused. This structural backstop ensures the author and the acquitting reviewer always come from different model families.

Are landed patches reversible?▼

Yes. Every target file is backed up to a dated backups directory before mutation, and each landing is logged with provenance metadata, so any change can be reverted from its backup.

Related Skills

Explore More Software Engineering Skills→
wanshuiyinwanshuiyin
15.6k

meta-apply

Applies approved self-modification patches to the skill corpus after cross-model jury review.

Community
Advanced
ronmikailovronmikailov

rollback-changes

Automate rollback of failed workflow phases using a validated changes log.

Community
Advanced
xqinagxqinag

meta-apply

Apply user-selected patches to the skill corpus after a fresh cross-model jury verdict.

Community
Advanced
czh-ee-2023czh-ee-2023

meta-apply

Approves and applies self-modification patches to the skill corpus after jury and human review.

Community
Advanced